COLLOQUIUM ON THE STATISTICAL SCIENCES
THE SCHOOL OF STATISTICS
University of the Philippines
invites all students and faculty to a colloquium on
Estimation of Multiple Time Series with Volatility
to be given by
Mark Louie F. Ramos
M.S. Statistics Candidate
University of the Philippines-Diliman
School of Statistics
Tuesday, 7 April 2015, 4:00 p.m.
UPSS Auditorium
Abstract
Volatility episodes in a multiple time series exert influence on the parameter estimates that will significantly reduce model fit for the non-volatile parts. Assuming a multiple time series with common autoregressive parameter, a combination of block bootstrap and forward search algorithm embedded into a backfitting algorithm was used to generate robust estimates when temporary volatility is present in the data.
Simulation studies exhibited robustness of the estimates from the hybrid algorithm. Furthermore, predictive ability of the fitted model is high during the non-volatile periods of the time series. While predictive ability deteriorates when the time series are very short or when they are nearly non-stationary, it still provides better estimates than some common modeling strategies.
Keywords: Multiple time series, volatility, modeling, bootstrap
COLLOQUIUM ON THE STATISTICAL SCIENCES
THE SCHOOL OF STATISTICS
University of the Philippines
invites all students and faculty to a colloquium on
CONSTRUCTION AND INTERNAL VALIDATION OF TWO-LEVEL NESTED LOGIT AND MULTIVARIATE PROBIT MODELS FOR FINANCIAL INCLUSION
to be given by
MR. MYNARD BRYAN R. MOJICA
Friday, 24 April 2015, 4:00 p.m.
UPSS Auditorium
ABSTRACT
The study constructed econometric models for financial inclusion based on two-level nested logit (NL) and multivariate probit (MVP) models. Results show that nesting the correlated outcomes of a polytomous response variable could improve the ability of the multinomial model in predicting an outcome with small choice probability. Simultaneous estimation of multiple correlated discrete response variables produced parameter estimates with lower standard error, and provided more similar predictive performance in training and test datasets. Internal validation of NL and MVP models was also performed using cross-validation and bootstrapping. For both models, internal validity is best estimated with bootstrapping which provided more stable estimates of prediction error with low bias.
Key words: multinomial models, nested logit, multivariate probit, cross-validation, bootstrapping
About the author:
COLLOQUIUM ON THE STATISTICAL SCIENCES
THE SCHOOL OF STATISTICS
University of the Philippines
invites all students and faculty to a colloquium on
Better Statistics: Key to Building Competitive Cities
to be given by
PROFESSOR OFELIA M. TEMPLO
Thursday, 30 April 2015, 4:00 p.m.
UPSS Auditorium
ABSTRACT
The Economics of Geography is a relatively new field in economics that highlight the role of cities in promoting economic growth at the national level, among others. Investment, in turn, is a critical factor in the attainment of economic growth, both at the national and local levels. The government has recognized this important linkage, promoting in the process local economic development thru a business-friendly and competitive environment for local government units. The United States Agency for International Development (USAID) is one of the development partners operating in the country that responded to the call of government to help create a competitive local environment especially in cities. Through the Investment Enabling Environment (or INVEST), USAID initiated several activities aimed at generating useful statistics that could be used for local policy making. These include the generation of the Cities and Municipalities Competitiveness Index and the Survey of Business Permits and Licensing Systems in the Philippines, both of which are being managed by the National Competitiveness Council Secretariat based at the Department of Trade and Industry. The Project also assisted in developing the eBPLS Readiness Survey released by the Department of Science and Technology to assess the potential of local governments in setting up computerized systems. At the local level, the INVEST Project also formulated an instrument for generating immediate feedback on the perception of business permit applicants which is conducted through an exit interview. As part of its contribution to the discussions on gender, the Project also analyzed the profile of business applicants in three cities, especially women-headed establishments. In the process of conducting these activities, it is clear that statistics or better data collection potentially play an important role in promoting investments and economic development in local government units, especially in cities.
About the author: